Technology

Researchers have developed a new type of light-emitting diode based on thin branched nanowires, which could offer significantly higher efficiency and lower production costs than current technology. By controlling where in the structure the light is generated, the researchers have reduced the losses that would otherwise limit the amount of light that can be utilised.

Research.

The results are in: Which AI model is the most fallible? Persuadable? Correctible?

Research assessed 7 different generative AI language learning models, or LLMs, for these three qualities during lengthy conversation. Their work reveals intrinsic limitations that might go undetected during one-off interactions.

Among the 7 LLMs tested – ChatGPT (GPT-3.5, GPT-4o, GPT-4o-mini), Claude 3.5, Sonnet, Gemini 1.5 Pro, Llama-3-70B, and DeepSeek-R1 – they found that:

• ChatGPT 3.5 was most vulnerable to reaffirming misinformation during a conversation containing repeated false statements; Claude 3.5 Sonnet was the least.
• All 7 were more susceptible to misinformation on obscure topics, implying that more training data on a given topic leads to more robust resistance to misinformation.
• DeepSeek was the most persuadable, as measured by responses to increasingly argumentative prompts, mostly because of its tendency toward sarcastic answers, which could not be reliably interpreted.
• 4 models – ChatGPT 4o, ChatGPT 4o-mini, Gemini 1.5 Pro, and DeepSeek – corrected errors 100% of the time when given a second opportunity.

A routine John Doe lawsuit could become part of Strike 3 Holdings' $446 million AI training case against Meta. The adult film producer wants to link the two lawsuits, arguing that a Reality Labs executive downloaded nearly 20,000 files at his home for work purposes, not for personal use. Meta counters that the alleged home downloads are not linked to the company.

WASHINGTON, Sept 4 (Reuters) - U.S. military officials say they have disabled advertising trackers on a range of phones and computers, according to letters, opens new tab released on Friday by U.S. Senator Ron Wyden and statements given ‌to Reuters, a development that follows reports that commercially available location data had been used to target American forces in the Middle East.

The disclosures highlight ⁠a growing national security concern: that location data collected by the advertising industry and sold by data brokers — companies that collate and resell personal data — can be used to track and target military personnel deployed to war zones.

Mayor Zohran Kwame Mamdani and Schools Chancellor Kamar H. Samuels today announced a moratorium on student-facing generative Artificial Intelligence (AI) use in schools, alongside a new screen time policy, a decision that will impact nearly 600,000 public school students, or two-thirds of the system’s total enrollment. The policy establishes a one-year moratorium, effective in the 2026-2027 school year, on student-facing generative AI for children in 2-K through 8th grade. It also introduces twice-yearly AI critical thinking modules for high schoolers, limited AI pilots for a small number of high school classrooms and age-appropriate screen time restrictions.

NVIDIA has agreed to acquire Hugging Face for $12,930,300,000. Together, we will scale Hugging Face’s platform, strengthen its infrastructure and expand access to AI for developers and institutions worldwide.

Today Electronic Privacy Information Center (EPIC), represented by Protect Democracy and Citizens for Responsibility and Ethics in Washington (CREW), filed a lawsuit challenging the Trump Administration’s unprecedented and illegal effort to interfere in the 2026 midterms by using unreliable databases for reasons totally unrelated to their intended purpose to create lists of “citizens” in every state for determining who can and cannot vote. The complaint details how the administration’s actions are not only illegal but also run the risk of disenfranchising untold numbers of eligible voters ahead of the midterms, unless the court intervenes.

On March 31, 2026, the White House issued its second executive order on elections, which directed the Department of Homeland Security (DHS), Social Security Administration, and the State Department to create “State Citizenship Lists” of “confirmed” citizens in all 50 states who reside in those states and are entitled to vote in federal elections. The data DHS is seeking to centralize—such as social security numbers, addresses, and citizenship information—is currently held by different federal agencies for data security and privacy purposes.

As directed by the executive order and a subsequent memo from DHS, states will have just 60 days to cross reference their voter rolls with the new federal State Citizenship Lists—presumably to purge or withhold ballots from voters suspected of being ineligible to vote. Alarmingly, DHS has admitted that these lists will contain widespread inaccuracies that could result in eligible voters being wrongly flagged or removed from state voter rolls just weeks before Election Day. In fact, states that have already used this data have reported widespread errors and voter disenfranchisement and warned that there is “no way those lists are accurate.” Nevertheless, in order to compel states to use their error-prone citizenship lists, the Trump administration is threatening to investigate and prosecute states and election officials who “issue Federal ballots to individuals not eligible to vote.”

To make matters worse, by consolidating Americans’ sensitive personal information (likely including Social Security numbers, dates of birth, and citizenship records) into a single federal system, the government is creating exactly the kind of centralized personal data repository that federal privacy law was designed to prevent—one that presents a significantly heightened risk of breach, misuse, and identity theft.